- OS: Ubuntu 16.04
- Language: Python3.5
For gpu version:
- Install cuda and cudnn. Their version should match tensorflow-gpu with version 1.8.0
- Install python package:
pip3 install -r requires.txt
For cpu version:
- Install python package:
pip3 install -r requires.txt
- Remove tensorflow-gpu and install tensorflow:
pip3 uninstall tensorflow-gpu & pip3 install tensorflow==1.8.0
- Generate all temporary file:
./build.sh
- It will generate a directory named
tmp
which contains all temporary files.
- Train model:
./train.py my_model
- It will spend about 25 minutes to train a model with 5-fold cross validation in a nvidia 1080ti GPU.
- Generate result:
./eval.py my_model
- It generate a directory named
result
. And there are two files in the directory.result.txt
: output file of this homeworkcheck.txt
: the file that show the sorted classification probability of each user